Researchers have developed a new framework for decoding speech from brain recordings taken without invasive equipment. The Cross-Subject Perceived Speech Decoding (CPSD) framework addresses the problem of limited generalizability across different subjects by using two stages: pre-training on multiple source subjects and fine-tuning on individual target subjects. A key component is the Positional Encoding-based Spatial Attention module, which standardizes brain data to improve model training. The approach was tested on three datasets with significant improvements over baseline methods. --- Why it matters: This matters because it could lead to more accurate and efficient speech recognition from non-invasive brain recordings, potentially benefiting applications such as assistive technology for people with paralysis or other motor disorders. Source: https://arxiv.org/abs/2608.22420